21 Tobacco: what a strong recommendation looks like
If you want to know whether this book is a hatchet job, this is the chapter to read.
Tobacco is the one subject in Delivering Better Oral Health where the underlying science is overwhelming. Smoking causes oral cancer. Stopping helps. Nobody serious disputes either claim, and the guideline is right to push it as hard as it does. I have no interest in manufacturing a complaint here.
What I want to do instead is apply the same method to a case where the answer comes out largely favorable, because a method that only ever finds fault is not a method. And then look at one row, because even here the guideline does the thing this book is about, and here it may be entitled to.
The advice
Tobacco gets seven recommendations across two tables, organized around a structure called Very Brief Advice: ask, advise, act. Six of the seven are Strong. Table 3a covers smoking and table 3b repeats the same three steps for smokeless tobacco.
The seventh is Conditional, it is about e-cigarettes, and an earlier draft of this chapter did not audit it at all. It turns out to be the most interesting row in the guideline, so it gets its own section below.
Ask. At every opportunity, ask patients if they smoke and record smoking status (smoker, ex-smoker, never smoker).
Strength of recommendation: Strong
Advise. Explain that a combination of behavioural support and the medication varenicline, or short-acting with long-acting Nicotine Replacement Therapy, are likely to be most effective.
Strength of recommendation: Strong
Act. Refer people who want to stop smoking to local stop smoking support, preferably where behavioural support and prescribed stop smoking medicines are available.
Strength of recommendation: Strong
— Delivering Better Oral Health, chapter 2, table 3a (1)
What the guideline says its evidence is
For the Act step:
Moderate certainty evidence. NICE guidance supports this recommendation as “evidence and expert opinion showed that support provided by these services is clinically effective and highly cost-effective in helping people to stop smoking”.
— Delivering Better Oral Health, chapter 13 (1)
That is a straightforward, well-supported, correctly labeled recommendation. Referral to stop smoking services works, it is cost-effective, and moderate certainty comfortably supports a Strong recommendation. There is nothing here for me to say except that it is right.
Now read the evidence statement for the Ask step.
No specific evidence regarding asking about smoking; however, it is an integral part of the Ask, Advise, Act pathway.
— Delivering Better Oral Health, chapter 13 (1)
A Strong recommendation, and the guideline’s own evidence statement opens by saying there is no specific evidence for it.
Is that a scandal?
No. And working out why not is more useful than pretending it is.
This is the clearest instance in the entire guideline of a component taking its strength from the structure it sits in rather than from evidence of its own. It is the pattern of Chapter 3, stated openly. But unlike the brushing bundle, here I think the borrowing is defensible, for three reasons that are worth spelling out because they are exactly what a Strong recommendation on no direct evidence has to demonstrate.
The component is logically necessary to the pathway. You cannot advise a smoker to quit, or refer them, without first establishing that they smoke. Asking is not a separate intervention competing for the patient’s attention; it is the entry condition for the interventions that do have evidence. Demanding a randomized trial of asking, against a comparator of not asking but somehow still advising, is demanding a trial of something incoherent.
The burden is trivial and the harm is close to nil. One question in a medical history. There is no plausible mechanism by which asking a patient whether they smoke makes their health worse, and nothing else is displaced by it.
The downstream effect is large. Brief opportunistic advice to stop smoking has a real effect on quit rates, and the population impact of applying it at scale is substantial because the underlying harm is enormous. When the thing downstream is that valuable, an entry step with no trial of its own inherits a strong case.
Those three conditions together are close to what GRADE has in mind when it allows a Strong recommendation despite low or absent certainty (2). Recall from Chapter 2 that Alexander and colleagues found only 15.6% of discordant WHO recommendations met that bar (3). This one does. I have criticized bundling repeatedly in this book; intellectual honesty requires saying clearly when the same move is justified.
What the guideline could still do, and does not, is say which of those three reasons it relied on. “No specific evidence, but it is integral to the pathway” is half of the argument. The half that is missing is the half that would let a reader check it.
The row I left out
Here is the row, in full.
Acknowledge that e-cigarettes may be helpful for some smokers for quitting or reducing smoking.
Strength of recommendation: Conditional
— Delivering Better Oral Health, chapter 2, table 3a (1)
And its evidence statement:
Recommendation based on low certainty evidence from one systematic review; insufficient evidence to demonstrate the long-term effects.
— Delivering Better Oral Health, chapter 13, table 15 (1)
That is an accurate summary of the review the guideline cites. The problem is which version of it.
The reference given is Hartmann-Boyce and colleagues, Electronic cigarettes for smoking cessation, Cochrane Database of Systematic Reviews 2016, article CD010216.pub3 (4). According to PubMed, that version searched to January 2016. It found 24 completed studies, of which three were randomized and only two contributed to the cessation meta-analysis, together enrolling 662 participants. It rated the evidence low for e-cigarettes against placebo e-cigarettes, very low against a nicotine patch, and concluded that “the long-term safety of ECs is unknown.”
So the guideline read its source correctly. It read a source that had already been revised several times.
The dates matter here and I got them wrong in an earlier draft, so Chapter 32 sets them out in full. The short version: the line “Updated 10 September 2025” at the top of every DBOH chapter records a formatting change. The last time the evidence was reviewed in full was 21 September 2021, published as the fourth edition. That is the date against which the panel should be judged.
| Version | Published | Search closed | Studies | Nicotine EC vs NRT |
|---|---|---|---|---|
.pub3 |
Sept 2016 | Jan 2016 | 24 | not assessed; low vs placebo EC |
.pub6 |
14 Sept 2021 | May 2021 | 61 | moderate, RR 1.53 (1.21 to 1.93), 4 studies, 1,924 |
.pub8 |
Jan 2024 | Jul 2023 | 88 | high |
.pub11 |
Aug 2026 | Jan 2026 | 80 RCTs | high, RR 1.61 (1.23 to 2.12), 11 studies, 4,114 |
At the fourth-edition review, the current version was .pub6, published on 14 September 2021, one week earlier (5). It reported moderate certainty that nicotine e-cigarettes increase quit rates compared with nicotine replacement therapy, from 61 studies overall and 4 trials in 1,924 participants for that comparison.
One week is not a fair deadline. But .pub4 appeared in October 2020 and .pub5 in April 2021, so a version more current than 2016 had been available for most of a year, and the gap being closed is five years wide.
Low and moderate are one step apart on GRADE’s four-level scale, which runs high, moderate, low, very low. The step still matters: it is the difference between “further research is very likely to change this estimate” and “further research is likely to have an important impact.” An earlier draft called low and high “the two ends of the GRADE scale”, which is wrong twice over, since very low sits below low and the version available in 2021 said moderate rather than high.
The part that is hard to explain
The review was not obscure, and it was not quietly revised. .pub6, the version sitting on the shelf the week the fourth edition appeared, already carried this sentence in its published abstract (5):
To ensure the review continues to provide up-to-date information to decision-makers, this review is now a living systematic review. We run searches monthly, with the review updated when relevant new evidence becomes available.
— Hartmann-Boyce and colleagues, CD010216 (5)
A living systematic review is one that is continuously updated as evidence appears, rather than revised every few years or not at all. The term has a methods literature behind it, and its motivating observation is precisely the one this section is making: reviews are hard to keep current, and “failure to do so leads to a decay in review currency, accuracy, and utility” (6).
So the situation is this. A UK national guideline supported a recommendation with a 2016 snapshot of a Cochrane review that searches for new evidence every month, that announces it does so in its own abstract, and that had moved a certainty level by the week the guideline was signed off.
It has kept moving since. The current version, .pub11, searched to January 2026 and reports high-certainty evidence that nicotine e-cigarettes beat nicotine replacement therapy: RR 1.61 (95% CI 1.23 to 2.12), from 11 trials in 4,114 participants, an additional four quitters per 100 (95% CI 1 to 7 more) (7). That version appeared while this chapter was being revised, and replaced the one an earlier draft called current. A chapter about evidence going out of date should probably expect that.
Why this one cuts the other way
Every other complaint in this book has the same direction: the label claims more than the evidence supports. This one runs backwards, and I want to be exact about it rather than fold it into the general grievance.
Here the guideline claims less. “May be helpful for some smokers” is the language of real uncertainty, and it was the right language for CD010216.pub3. Against the review as it stood when DBOH went to press, it understates what is known. A dental team reading table 3a would take away that e-cigarettes are a marginal option of unclear benefit. The review they were pointed at says, at high certainty, that nicotine e-cigarettes outperform the nicotine replacement therapy the guideline recommends two rows above.
I am not saying the label should have been Strong. Strength is an evidence-to-decision judgment, and the panel is entitled to weigh long-term safety, youth uptake, and the acceptability of recommending a nicotine product in a dental setting. The review itself is still cautious about serious adverse events, where the confidence intervals remain wide. Those are legitimate reasons to stay Conditional on high-certainty benefit, and GRADE has a name for that situation.
The criticism is narrower and it is not about the strength label at all. It is that the certainty statement is wrong on the guideline’s own terms. Chapter 13 tells the reader “low certainty evidence from one systematic review.” The systematic review it means said “high certainty” eight months before publication. A reader has no way to detect this, because chapter 13 gives the certainty but not the version, and the reference list gives the version but not the certainty. You have to hold both pages open at once, which is what this book is for.
What the dental evidence actually shows
There is one further thing worth surfacing, and it is not a criticism of the recommendation so much as a correction to how it is often described.
The Cochrane review of tobacco cessation delivered by dental professionals includes 20 trials in 14,897 participants (8). According to PubMed, its findings are:
| Intervention | Effect on abstinence at ≥6 months | Certainty |
|---|---|---|
| Behavioral support, one session | RR 1.86 (95% CI 1.01 to 3.41), I²=66%, 4 studies | very low |
| Behavioral support, more than one session | RR 1.90 (95% CI 1.17 to 3.11), I²=61%, 7 studies | very low |
| Behavioral support plus NRT or e-cigarettes | RR 2.76 (95% CI 1.58 to 4.82), I²=0%, 4 studies | moderate |
| Behavioral support in schools or colleges | RR 1.51 (95% CI 0.86 to 2.65), I²=83%, 3 studies | very low |
Sixteen of the twenty trials were judged at high risk of bias, three at low, one unclear. The review’s own conclusion is that quit rates increase with behavioral support at very low certainty, and that abstinence increases at moderate certainty when pharmacotherapy is added.
Two things follow.
The strongest result in the table is the one that involves medication. The only moderate-certainty finding, and by some distance the largest effect, is behavioral support combined with NRT or e-cigarettes, with no heterogeneity at all (I²=0%). That is a striking result and it maps exactly onto what the guideline tells dental teams to do: refer to services “preferably where behavioural support and prescribed stop smoking medicines are available.” The recommendation and its best evidence line up. This is what the previous chapters kept failing to find.
Behavioral support delivered by dental teams alone is very low certainty. This matters for how the profession talks about itself. The claim “dentists can help patients quit smoking” is supported; the claim “this has been robustly demonstrated for dental settings specifically” is not, and the review says so. The absolute numbers are modest: for a single session, roughly 45 quitters per 1,000 against 24 per 1,000.
PubMed, 21 August 2026. The Cochrane review’s own search ran to February 2020 across the Cochrane Tobacco Addiction Group’s Specialised Register; I have not attempted to reproduce it. I checked for a .pub5 update and found none, so CD005084.pub4 remains current.
Provenance. My first attempt to obtain this review failed: Cochrane deposits into PubMed Central sit outside its Open Access subset, so the automated route returned nothing, and an earlier draft of this chapter said the full text was unavailable to me. It was later supplied, and both the review and its characteristics-of-studies tables are now in documentation/refs/. The numbers reported above match the full text. Where I report that sixteen studies were at high risk of bias, that is still the review’s own summary statement rather than my independent re-rating of its domain judgments, and I flag the difference because the two are not the same thing.
Second search, on the e-cigarette row. Added later, after this chapter was found to have skipped DBOH-074 entirely. PubMed, for all versions of CD010216. I retrieved metadata for .pub3 (2016), .pub6 (September 2021), .pub8 (2024), .pub9 (January 2025), .pub10 (November 2025) and .pub11 (August 2026) and read their structured abstracts. Version numbers, search closing dates, study counts, effect estimates and certainty ratings in Table 21.1 are taken from those abstracts. I did not retrieve .pub4, .pub5 or .pub7; their dates come from the PubMed record rather than from reading them.
Dates corrected. An earlier version of this section judged the panel against 10 September 2025, which is the date DBOH’s pages were last reformatted rather than reviewed. The comparison is now against the fourth-edition evidence review of 21 September 2021, and the claim is correspondingly narrower: the version available then reported moderate certainty, not high.
Full record in appraisals/searches/ch21-tobacco.md.
Verdict
- Certainty of evidence
-
Mixed, and correctly described by the guideline. Moderate for referral to stop smoking services and for behavioral support combined with pharmacotherapy. Very low for behavioral support delivered by dental professionals on its own. None at all for the asking step, as chapter 13 states in terms.
Except for e-cigarettes, where the guideline is out of date rather than wrong. Chapter 13 reports low certainty, quoting a 2016 version of a review that had reached moderate certainty for nicotine e-cigarettes against nicotine replacement therapy by the week the fourth edition was published, and high certainty since.
- Directness to the advice as worded
- Good. The cited evidence is about smoking cessation, in the settings and populations the advice addresses, measuring abstinence at six months or more. That is a patient-important outcome measured directly, which is more than most chapters of this book can say.
- Is the strength label defensible?
-
Yes, including for the Ask step with no evidence behind it. It meets what GRADE requires of a discordant recommendation: the component is logically necessary to a pathway whose downstream steps are well supported, the burden is a single question, and no plausible harm attaches to it. This is the category that Chapter 2 said exists and is rare, and here it applies.
The criticism that survives is smaller and is about drafting, not judgment: the guideline states that there is no specific evidence and that the step is integral to the pathway, but does not set out why that combination justifies Strong. A reader is asked to accept the conclusion without the reasoning.
Conditional for e-cigarettes is also defensible, on residual uncertainty about serious adverse events and long-term use, which is a values judgment the panel is entitled to make. What is not defensible is the certainty statement attached to it, which describes an evidence base that had already been replaced. See Chapter 32.
- What would change my mind
- For the Ask step, evidence that recording smoking status displaces something more valuable in a short appointment, or that it deters attendance among people who smoke. For the dental-delivered behavioral support, better trials: the review asks specifically for biochemical validation of abstinence rather than self-report, and sixteen of twenty trials at high risk of bias is a literature with a great deal of room to improve. For the e-cigarette row, nothing about the evidence: it is the guideline’s description of the evidence that needs changing, not the evidence.
What this means for you
If you smoke, the single most useful thing in this entire book is on this page, and it is not about your teeth. Stopping smoking is worth more to your health than every other recommendation in Delivering Better Oral Health put together.
If your dentist asks whether you smoke and it feels intrusive or repetitive, it is the entry point to the one intervention here with a large effect on mortality. And if you want to stop, the evidence points clearly at one thing: behavioral support combined with medication, through a stop smoking service, rather than willpower alone or advice alone.
On vaping, the guideline is more lukewarm than the evidence it points to. Its own cited review, in the version current when it was published, found at high certainty that nicotine e-cigarettes help more people stop smoking than nicotine patches or gum do. If you are choosing between them, that is worth knowing. The open question is long-term safety, where the trials are still too small and too short to settle anything, so the honest summary is that switching completely from smoking to vaping is well supported as a way to stop smoking and is not the same as being proven harmless.